Predictors of serum testosterone and DHEAS in African-American men

Summary

There are few reported data on biochemical and functional correlates of androgen levels in African-American men. This study aimed at reporting physical and biochemical correlates of serum total testosterone (total T), bioavailable testosterone (BT) and dehydroepiandrosterone-sulphate (DHEAS) levels in community-dwelling, African-American men aged 50–65 years. Home-based physical examinations and health status questionnaires were administered to randomly sampled men. Body composition (dual-energy X-ray absorptiometry), lower limb and hand-grip muscle strength, and neuropsychological functions were assessed. Levels of serum total T, BT, DHEAS, oestradiol (E2), adiponectin, leptin, triglycerides and glucose were measured. Multiple linear regression models were constructed to identify factors independently associated with androgen levels. DHEAS levels declined from age 50 to 65 years (p < 0.0001), but total T and BT levels remained constant. Independent of other associated factors, higher total T levels were associated with lower serum triglyceride levels (β = −0.142, p = 0.049); higher BT was associated with better performance on the trail-making tests (TMT-B:TMT-A ratio: β = −0.118, p = 0.024) and higher DHEAS levels were associated with lower adiponectin (β = −0.293, p = 0.047) and higher mini-mental state examination (MMSE) score (β = 0.098, p = 0.008). Multiple regression models predicted 21, 18 and 29% of variance in total T, BT and DHEAS, respectively. Higher total T levels were associated with serum metabolic markers, particularly lower triglycerides, whereas higher BT was associated with better cognitive and muscle function and DHEAS with lower adiponectin and higher MMSE scores

Introduction

In men, bioavailable testosterone (BT) and free testosterone (FT) levels decline by about 1.0 and 1.2% per year, respectively, and total testosterone (total T) declines by about 0.4% per year after the age of 40 years (Morley et al., 1997; Feldman et al., 2002). Serum levels of both the testosterone precursor, dehydroepiandrosterone (DHEA) and its active sulphated form (DHEAS) also decline with increasing age. The definition of clinically relevant androgen deficiency in the ageing male remains uncertain. Clinical features common to both ageing and androgen deficiency include decreased muscle mass and strength, increased fatigue, increased fat mass, loss of libido, erectile dysfunction, impaired cognitive function and depression. It is, however, difficult to separate the effects of ageing per se as opposed to a range of other concomitant health-related factors on plasma testosterone (Haren et al., 2002). It is still unclear, after many cross-sectional and a number of longitudinal analyses, as to which testosterone measures [total T, BT, FT (measured or calculated)] best correlate with symptoms typical of androgen deficiency in the older male.

Associations between serum testosterone levels and function in adult men are being increasingly described; however, few data exist specifically on African-American men. Low levels of gonadal steroids and changes in insulin-like growth factor (IGF) axis activity may be markers of the metabolic syndrome associated with visceral obesity and loss of muscle mass (sarcopenia) (Morley et al., 2005). The strongest predictor of sarcopenia as found in the New Mexico Ageing Process Study (Baumgartner et al., 1999) was FT. Age, caloric intake, physical activity and IGF-1 levels were also predictive. Sarcopenia leads to frailty [defined by weight loss, exhaustion, weakness (grip strength), slow walking speed and low physical activity; Fried et al., 2001], an important precursor of subsequent functional deterioration and death (Newman et al., 2001; Baumgartner et al., 2004). Obesity, particularly visceral, is associated with low plasma testosterone levels (Zumoff et al., 1990; Khaw & Barrett Connor, 1992; Tchernof et al., 1997). Therefore, low levels of gonadal steroids and changes in the activity of the IGF axis may be markers of the metabolic syndrome associated with visceral obesity, sarcopenia, increased frailty, functional deterioration and death.

Even in otherwise healthy people, a variable age-related decline in cognitive function occurs in the domains of processing speed, reasoning, memory and spatial abilities (Verhaeghen & Salthouse, 1997). Individual differences in the rate and severity of decline are likely to be due, in part, to variance in health status, socioeconomic status, lifestyle and genetics. Testosterone has been identified as a potentially modifiable factor in both normal and pathological cognitive ageing in males (Flood et al., 1995; Morley et al., 1997). A number of population-based studies of ageing community-dwelling men have reported positive associations between testosterone levels and neuropsychological domains such as learning and memory (Morley et al., 1997; Barrett-Connor et al., 1999; Moffat et al., 2002), processing speed (Yaffe et al., 2002; Muller et al., 2005) and executive function (Barrett-Connor et al., 1999; Muller et al., 2005).

In the present cohort of African-American men sampled from Inner City and suburban St Louis, Missouri, it has been reported that physical and neuropsychological disability is very common (Miller et al., 2004, 2005). In this study, we examined the functional and biochemical correlates of serum total T, BT and DHEAS levels in this group of community-dwelling, African-American men aged 50–65 years.

Materials and methods

Population

This group represented a sub-group of a wider cohort of African-American men and women, sampled from two regions: Inner City and suburban St Louis, Missouri. The African-American Health (AAH) study is a population-based longitudinal study of 998 African-Americans (371 men and 627 women) from diverse socioeconomic areas of St Louis, Missouri to address the disparity in disabilities among middle-aged African-Americans. [The sampling and data collection procedures have been previously described in detail (Miller et al., 2004, 2005).] Baseline assessments occurred between 2000 and 2001.

To be included in the study, participants had to self-report black or African-American race, be born between 1936 and 1950, have an MMSE of ≥16 and provide written informed consent. The sub-group analysed for the present study included 124 men with a mean age of 56.1 ± 4.4 years (range 50–65) and body mass index (BMI) of 28.54 ± 5.25 kg/m² (range 17.76–43.37), who agreed to participate in additional studies which included blood sampling. This study was approved by the Saint Louis University Institutional Review Board (IRB).

Physical examination, interview and questionnaires

A home-based physical examination and a health status questionnaire were used to record co-morbid disease. The questionnaire included age, environmental characteristics and annual income. Chronic conditions [hypertension, diabetes mellitus, angina, chronic heart failure (CHF), coronary artery disease (CAD), stroke] were assessed by asking the participant ‘Did a doctor ever tell you that you have…?’ All medications, prescription and over-the-counter, were recorded by the interviewer during the in-home assessment. Medications were categorized as oral anti-diabetic, anti-hyperlipidaemic, anti-hypertensive (further categorized as angiotensin-converting enzyme inhibitor, angiotensin II receptor blocker, anti-adrenergic, beta-blocker, calcium channel blocker and diuretic), androgens and thyroxine. Data on smoking were obtained by direct inquiry.

Hip circumference was determined by using a tape measure at the level of the maximum posterior protrusion of the buttocks. Waist circumference was measured 1 cm above the iliac crests. Waist-to-hip ratio (WHR) was calculated. BMI was calculated based on the measured height and weight (kg/m²).

Body composition

Total body lean mass (LM) and fat mass (FM), the composition of the appendicular skeletal mass (ASM) and bone mineral density (BMD) of cervical spine, hip and lumbar spine were evaluated by dual-energy X-ray absorptiometry (DEXA) (Hologic QDR 4500 W Bedford, MA).

Muscle strength and physical function

Isokinetic strength testing of lower limbs (flexion and extension at ankle, knee and hip) was performed using Biodex System 3 Pro (Biodex Medical Systems, Shirley, NY, USA). Peak torque, work, power, time-to-peak torque and the angle of the joint at which peak torque was obtained were assessed through 60° and 120° of motion. Data were corrected for body weight. Only peak torque data generated through 120° of motion were used in these analyses and are reported in Newton-metre (N m).

Isometric grip strength was assessed using a handgrip dynamometer (Jamar, Jackson, MI, USA). The mean of three maximal effort trials with the self-reported strongest hand was used in these analyses. The test was performed with the subject seated in a chair (without arm rests), with feet flat on the floor and the strongest arm held flat against the side with the elbow at 90°.

The sit–stand test measured the time taken to accomplish five complete raises from a chair. The gait speed test measured the average time (two trials) taken to walk a 4-m course at a usual pace, as if walking to the store. The 6-min walk test measured the distance (m) walked in the allotted 6-min time period on a 90-ft pathway.

Activites of daily living (ADLs) were assessed (Lawton & Brody, 1969). These consisted of seven basic ADLs (BADLs: having any difficulty with bathing, dressing, eating, getting into and out of bed or chairs, walking across a room, getting outside, and using the toilet; potential range 0–7) and eight instrumental ADLs (IADLs: having any difficulty with preparing meals, shopping for groceries, managing money, making telephone calls, doing light housework, doing heavy housework, getting to places outside walking distance, and managing medications; potential range 0–8). Physical performance scores (Nagi, 1976) were assessed for lower body (difficulties in walking a quarter of a mile, walking up and down 10 steps without rest, standing for 2 h, stooping, lifting 10 pounds, and pushing large objects; range 0–6) and upper body (difficulties reaching up over one’s head, reaching out as if to shake hands, or grasping; range 0–3). Physical activity levels were assessed by using the Yale Physical Activity Study (YPAS) scale (Dipietro et al., 1993).

Neuropsychological function

The MMSE is a tool used to detect cognitive deficits seen in syndromes of dementia and delirium and for measuring these cognitive changes over time (Cockrell & Folstein, 1988). A score of ≤23 is 87% sensitive and 82% specific in detecting dementia and delirium when a psychiatrist’s standardized clinical diagnosis is used as the criterion (Anthony et al., 1982). Depressive symptoms were assessed by using the 11-item Center for Epidemiologic Studies Depression Scale (CESD) (Kohout et al., 1993; Miller et al., 2004).

The trail-making test (TMT) is a test of visuomotor tracking and attention (Greenlief et al., 1985). Crowe (1998) demonstrated that performance on part A of the TMT was uniquely determined by visual search and motor speed and that performance on part B was determined by lowered reading level, poor skill in visual search, poor ability to maintain two simultaneous sequences and decreased attention and working memory functions. Part A of the test consists of circled numbers 1–25 on a page. Participants were first given a sample trail to ensure that they understood the task. They traced a line beginning at circle 1 and ending at circle 25 in as little time as possible. If errors were made they were immediately asked to restart from the last correct move. Part B of the test consists of numbers 1–13 and letters A–L on a page. Participants were required to draw a specific trail between these circles alternating sequentially between numbers and letters in as little time as possible. The TMT is reported as the ratio of part B:part A (TMT-B:TMT-A), an index of executive function (Arbuthnott & Frank, 2000), where lower scores indicate better executive function.

Laboratory analyses

Blood was drawn for laboratory analyses at the time of the DEXA examination. Thus, they represent neither morning nor fasting values. Total T was measured by radioimmunoassay (RIA, ICN-Biomedicals, Costa Mesa, MA) with intra- and inter-assay coefficients of variation (CVs) of 6.7 and 7.3%, respectively. BT was assessed by the ammonium sulphate precipitation method (O’Connor et al., 1973). The intra- and inter-assay CVs were 7.5 and 10.1%, respectively. Adiponectin was determined using a commercially available RIA kit (Linco Research, St Charles, MO, USA), with intra- and inter-assay CVs of 5.3 and 8.1%, respectively. Serum E2 and leptin were measured by RIA kit (Diagnostic Systems Laboratories, Webster, TX, USA). The intra- and inter-assay CVs were 6.5 and 9.7% for E2 and 4.7 and 5% for leptin, respectively. Serum DHEAS was measured by RIA (Diagnostic Products Corp., Los Angeles, CA) with intra- and inter-assay CVs of 5.3 and 7.0%, respectively. Triglycerides were measured using a commercially available enzymatic kit from Roche Diagnostics (Indianapolis, IN, USA). In this study, triglycerides had an intra-assay CV of 1.1% and inter-assay CV of 3.6%. Serum glucose measurement was performed in a commercial clinical laboratory (Smith Kline-Beecham, St Louis, MO, USA).

Statistical analysis

Serum total T, BT and DHEAS levels were available for 123 male participants. Most hormonal variables presented a log-normal distribution. Total T, BT, DHEAS, E2, adiponectin, leptin, triglycerides and glucose were normalized by transformation into their natural logarithm. Linear regression (continuous variables) in combination with correlation analyses, was initially used in the whole population to detect potential predictors of serum total T, BT and DHEAS levels (124 individual statistical tests). For dichotomous variables (medication use, chronic health conditions and smoking), the differences in hormone levels between the groups were assessed using two-tailed t-tests (60 individual statistical tests). Variables significantly associated with hormone levels at an alpha level of 0.05 were further included together in multiple regression predictive models. Multiple regression data were presented as the predictive value of the model (adjusted R²) and associated p-value. The impact of individual factors was expressed as a β coefficient and associated p-value. Additionally, two-tailed t-tests were used to analyse the mean differences in outcome measures between men with biochemically defined hypogonadism (BT < 70 ng/dL) and eugonadal men, defined by BT >100 ng/dL (58 individual statistical tests). Data analyses were performed using Intercooled Stata 7.0 (STATA Corporation, College Station, TX, USA). Fig. 1 was generated using GraphPad Prism 4.0 (GraphPad Software Inc., San Diego, CA, USA).

Figure 1.

The effect of age on serum hormones in African-American aged 50–65 years. Only log DHEAS declined significantly in men between 50 and 65 years (R = −0.366, p < 0.0001). T, BT and E2 levels remained constant (dashed line, p > 0.05). DHEAS, dehydroepiandrosterone; T, total testosterone; E2, oestradiol; BT, bioavailable testosterone.

Results

Strata, income, age and hormones

Serum total T and BT levels were significantly lower in suburban-dwelling men when compared with inner city men (p = 0.012 and 0.015, respectively). There were no significant differences in age, BMI, waist circumference, systolic or diastolic blood pressure, DHEAS or E2 between men residing in the city compared with those in the suburban area (Table 1).

Table 1.

Cohort characteristics by strata (inner city/suburban)

	Mean	SD	Min.	Max.	N
Inner city
Age (years)	56.22	4.63	50	64	55
T (ng/dL)	486.65	193.94	219	1119	54
BT (ng/dL)	104.69	47.28	42	284	54
DHEAS (μg/dL)	119.67	83.01	3	348	55
E2 (pg/mL)	19.53	7.49	4	40	55
SBP (mmHg)	135.06	28.53	74	222	52
DBP (mmHg)	84.31	13.12	60	117	52
BMI (kg/m2)	28.48	6.31	17.76	43.37	48
Waist (cm)	101.87	15.46	78.1	133.5	26
Suburbs
Age (years)	55.99	4.19	50	65	69
T (ng/dL)	410.58*	182.4	121	1114	69
BT (ng/dL)	88.26*	46.86	27	267	69
DHEAS (μg/dL)	114.04	71.82	4	326	69
E2 (pg/mL)	21.06	10.49	5	76	69
SBP (mmHg)	137.78	23.24	86	211	69
DBP (mmHg)	85.67	12.63	66	120	69
BMI (kg/m2)	28.58	4.29	19.79	40.78	61
Waist (cm)	99.69	11.79	78.75	127.2	34

^*p < 0.05 (comparison of log hormone values). There were no significant differences in age, DHEAS, E2, SBP, DBP, BMI or waist circumference between strata.

Data presented are mean ± SD and minimum and maximum values.

T, total testosterone; BT, bioavailable testosterone; DHEAS, dehydroepiandrosterone sulphate; E2, oestradiol; SBP, systolic blood pressure; DBP, diastolic blood pressure; BMI, body mass index.

DHEAS declined significantly over the 15-year age span (p < 0.0001), but total T, BT and E2 did not (Fig. 1). Total T was strongly correlated with BT (R = 0.53, p < 0.0001) and DHEAS was weakly correlated with E2 (R = 0.23, p < 0.05). There were no other significant correlations among the hormone levels.

Income level was inversely associated with total T levels (R = −0.27, p = 0.002) and there were trends towards an inverse association with BT (R = −0.16, p = 0.072) and a positive association with DHEAS levels (R = 0.18, p = 0.051).

Associations with hormones

Blood chemistry

Table 2 describes the unadjusted associations of total T, BT and DHEAS with biochemical, morphological and functional variables. Total T levels were positively associated with adiponectin levels (p = 0.03) and inversely associated with triglycerides (p = 0.003), leptin (p = 0.0002) and glucose levels (p = 0.005). BT levels were inversely associated with glucose (p = 0.034), but were not associated with adiponectin, triglycerides or leptin levels. DHEAS levels were inversely associated with adiponectin levels (p = 0.0001), but were not associated with triglycerides, leptin or glucose levels.

Table 2.

Associations between serum androgens and biochemical, physical and neuropsychological variables

	log T		log BT		log DHEAS
	R	p-value	R	p-value	R	p-value
Biochemical
Log adiponectin	0.2	0.03	0.032	0.733	−0.350	0.0001
Log triglycerides	−0.27	0.003	−0.076	0.411	−0.096	0.298
Log leptin	−0.33	0.0002	−0.088	0.346	0.143	0.124
Log glucose	−0.25	0.005	−0.194	0.034	−0.069	0.454
Physical
Systolic BP	0.054	0.56	0.009	0.925	0.069	0.45
Diastolic BP	0.004	0.97	−0.043	0.642	0.131	0.154
BMI	−0.27	0.003	−0.168	0.065	0.148	0.103
Waist circumference	−0.24	0.062	−0.188	0.151	0.049	0.713
Max. grip strength	−0.095	0.32	0.025	0.791	0.152	0.11
BADLs	0.146	0.108	0.068	0.452	−0.056	0.343
IADLs	0.147	0.107	0.166	0.068	−0.063	0.488
Nagi upper body	0.175	0.054	0.084	0.357	−0.079	0.382
Nagi lower body	0.101	0.267	0.135	0.137	−0.145	0.109
YPAS	0.039	0.67	0.103	0.258	0.08	0.377
Sit–stand test	−0.062	0.68	−0.128	0.385	0.04	0.787
6-m walk time	0.011	0.94	0.084	0.56	0.062	0.668
6-min walk distance	0.005	0.97	0.021	0.885	0.125	0.383
Peak torque/body weight (120°)
Ankle extensors	0.162	0.233	0.148	0.276	0.132	0.334
Ankle flexors	0.123	0.366	0.127	0.35	0.068	0.618
Knee extensors	0.125	0.351	0.231	0.081	0.161	0.226
Knee flexors	0.114	0.393	0.166	0.213	0.056	0.678
Hip extensors	0.262	0.06	0.297	0.033	0.03	0.833
Hip flexors	0.056	0.69	0.053	0.711	−0.115	0.419
Total skeletal mass
Fat mass/height	−0.311	0.016	−0.179	0.172	0.103	0.433
Lean mass/height	−0.212	0.104	−0.05	0.707	−0.070	0.597
Appendicular skeletal mass
Total	−0.29	0.023	−0.064	0.622	0.089	0.493
Percent lean	0.25	0.052	0.183	0.159	−0.080	0.539
Bone mineral density
Cervical spine	−0.09	0.49	0.076	0.56	0.036	0.786
Trochanter	−0.09	0.48	−0.001	0.994	−0.272	0.036
Intertrochanter	−0.12	0.355	0.005	0.971	−0.129	0.326
Total hip	−0.109	0.409	−0.012	0.925	−0.198	0.129
Lumbar spine	−0.098	0.458	0.108	0.413	0.148	0.26
Neuropsychological
MMSE	−0.077	0.4	0.0005	0.995	0.324	0.0002
TMT-A	−0.07	0.62	−0.163	0.242	0.051	0.714
TMT-B	−0.077	0.59	−0.339	0.015	−0.008	0.955
TMT-B:TMT-A	−0.073	0.61	−0.313	0.025	−0.093	0.515
CESD	0.013	0.083	0.014	0.089	−0.008	0.607

Data presented are Pearson’s correlation coefficients (R) and associated p values.

T, total testosterone; BT, bioavailable testosterone; DHEAS, dehydroepiandrosterone sulphate; BADLs, basic activites of daily living; IADLs, independent activities of daily living; YPAS, Yale Physical Activity Scale; MMSE, Mini-Mental State Examination; TMT, trail-making test; CESD, Center for Epidemiological Studies Depression Scale. Bold text indicates significant associations (p < 0.05); italicized text indicates trends of p < 0.07.

Body composition

Total T levels were inversely associated with BMI (p = 0.003), height-corrected fat mass (p = 0.016) and total ASM (p = 0.023), but BT levels were not associated with any measure of body composition. DHEAS levels were inversely associated with BMD at the trochanter (p = 0.036), but were not associated with any other measure of body composition.

Muscle strength and physical function

Bioavailable testosterone levels were positively associated with maximal peak torque generation in the hip extensor muscles (p = 0.033). Total T and DHEAS levels were not significantly associated with any measures of muscle strength or physical function.

Neuropsychological function

Bioavailable testosterone levels were inversely associated with the time taken to complete TMT-B (p = 0.015) and with the TMT-B:TMT-A ratio (p = 0.025). DHEAS levels were positively associated with the total MMSE scores (p = 0.0002). Total T levels were not associated with CESD scores, MMSE or TMT scores.

Smoking, chronic health conditions and medications

Total T and DHEAS levels were significantly higher in current smokers when compared with non-smokers (p = 0.033 and 0.037, respectively, data not shown). There were no significant effects of chronic health conditions or medication use on total T levels. BT levels were significantly lower in men taking anti-diabetic medication (p = 0.032, data not shown) and DHEAS levels were significantly lower in men taking beta-blocker medications (p = 0.008, data not shown).

Predictive modelling of hormones

Total T

The predictive model for total T (strata, income, adiponectin, triglyceride, leptin, glucose, BMI, current smoking), significantly predicted 21% of the variation in serum T levels in the sample (R² = 0.21, p = 0.0001). Only serum triglyceride levels (β = −0.142, p = 0.049) remained independently associated with total T in the model. There was also a trend towards an independent inverse association with income (β = −0.031, p = 0.053).

BT

The predictive model for BT (strata, glucose, hip extensor peak torque, TMT-A:TMT-B and anti-diabetic medication), significantly predicted 18% of the sample variance in serum BT levels (R² = 0.18, p = 0.023). The TMT-B: TMT-A ratio was the only factor that demonstrated an independent association with serum BT levels (β = −0.118, p = 0.024).

DHEAS

The predictive model for DHEAS (age, E2, adiponectin, trochanter BMD, MMSE, current smoking, beta-blocker medications), significantly predicted 29% of the sample variance in serum DHEAS levels (R² = 0.27, p = 0.0007). Serum adiponectin level was independently inversely (β = −0.293, p = 0.047) and total MMSE score was independently positively (β = 0.098, p = 0.008) associated with serum DHEAS levels.

Post hoc analysis (hypogonadal vs. eugonadal men)

In terms of defined hypogonadism (BT <70 ng/dL in this study), maximal hip (24.0 ± 3.71 vs. 33.91 ± 2.52 N m, p = 0.03) and knee extensor peak torque (44.13 ± 4.47 vs. 58.16 ± 4.67 units, p = 0.037) were lower in hypogonadal men when compared with men with BT >100 ng/dL, and there was a trend towards lower peak torque generation in ankle extensors (14.36 ± 1.92 vs. 20.06 ± 2.47 N m, p = 0.077). Moreover, mean TMT-B times (136.17 ± 14.53 vs. 89.04 ± 9.37 s, p = 0.009) and TMT-B:TMT-A ratios (3.42 ± 0.31 vs. 2.39 ± 0.21 s, p = 0.01) were higher in hypogonadal men when compared with men with BT >100 ng/dL. There were also trends towards lower serum glucose levels [4.79 ± 0.059 vs. 4.65 ± 0.053 mg/dL (log), p = 0.079] and more anti-adrenergic medication use (χ² = 3.77, p = 0.052) in hypogonadal men when compared with men with BT >100 ng/dL.

Discussion

Over the 15-year age-span covered by the present cross-sectional study, DHEAS levels declined significantly with age. However, there was no significant association between age and either total T or BT. This provided a unique opportunity in the context of current cross-sectional data on testosterone in older men, to investigate associations without the need to adjust for age.

Functionally, this cohort has been previously described as possessing a very high burden of both physical and neuropsychological disability Miller et al. (2005). Poorer executive function as assessed by the ratio of TMT-B:TMT-A was a significant independent predictor of lower serum BT levels. Moreover, higher scores on the MMSE independently predicted higher serum DHEAS in the present cohort. The epidemiological literature on associations between androgen levels and cognitive function in ageing men is somewhat confusing and disagreement between studies is related, in part, to differences in the tools used to assess the various domains of cognitive function. Higher serum BT but not total T levels were associated with better test scores on the TMT-B, digit symbol and MMSE in men with a mean age of 73 years (Yaffe et al., 2002). Testosterone appears particularly to improve spatial and frontal lobe-mediated working memory (Janowsky et al., 1994, 2000). These findings are in agreement with BT correlating with longitudinal memory decline in older males (Morley et al., 1997; Barrett-Connor et al., 1999).

Androgens may have a permanent organizing effect on some cognitive abilities because men with the idiopathic form of hypogonadism have markedly impaired spatial ability that does not improve with testosterone therapy, whereas visuospatial ability in men with acquired hypogonadism is similar to controls (Hier & Crowley, 1982). However, exogenous testosterone administration to older men increased visuospatial and memory test performance in some (Janowsky et al., 2000; Cherrier et al., 2001; Kenny et al., 2002), but not all studies (Sih et al., 1997; Haren et al., 2005). Testosterone improves memory in the SAMP8 mouse, an animal model of amyloid-beta protein overproduction, as does DHEAS (Flood et al., 1995; Flood & Morley, 1998; Farr et al., 2004). In tissue culture, testosterone reduces the production of amyloid precursor protein (Pike, 2001). Low testosterone, perhaps particularly FT levels are associated with the future development of Alzheimer’s disease (Hogervorst et al., 2001; Moffat et al., 2004).

Forty-three of the 123 men (approximately 35%) had BT levels of ≤70 ng/dL (an historical cut-off used in this laboratory, equivalent to 2 SD below the young adult mean). In the present study, hypogonadal men had lower muscle strength in both hip and knee extensor muscle groups and the failure of ankle extensor strength to make statistical significance may be related to sample size, measurement variance and statistical power. Much of the data relating to muscle function in the elderly are confounded by wide variability of the measures used (Clague et al., 1999). An analysis of the New Mexico Ageing Process Study by Baumgartner et al. (1999) demonstrated that in older males the best predictor of loss of muscle mass and strength was calculated free testosterone. In the present study, serum BT levels were positively associated with maximal peak torque generation in hip extensor muscles; this association was not independent of strata, glucose levels or anti-diabetic medication.

Muscle mass decreases and fat mass, particularly visceral fat, increases as men age and an inverse relationship has been described between testosterone and sex hormone-binding globulin (SHBG) at baseline and central adiposity 12 years later as estimated by WHR (Khaw & Barrett-Connor, 1992). Decreased testosterone levels in men are associated with increased accumulation of visceral fat (Tchernof et al., 1996; Marin & Arver, 1998), which is reversible upon testosterone administration (Lovejoy et al., 1995; Marin & Arver, 1998). Testosterone has been shown to increase lean mass and decrease fat mass after 6 months of treatment in men aged over 60 years with total T levels in the low to normal range (Wittert et al., 2003). Serum testosterone was inversely associated with BMI and height-corrected fat mass in the present study, and there was also a trend towards an association between total T and the percentage of lean tissue in the appendicular skeletal mass. However, these associations were weak and did not contribute strongly to the explanation of variance in testosterone levels in the cohort.

Studies reporting associations between circulating androgen levels and bone mineral density (BMD) at various skeletal sites have yielded conflicting results. After adjusting for age and BMI, hip BMD was positively correlated with the free androgen index (FAI) in 134 elderly men (Murphy et al., 1993) and inversely correlated with SHBG in 12 men with idiopathic osteoporosis and 12 normal men (Gillberg et al., 1999). Kenny et al. (1998) reported an absence of relationships between BMD and BT and FT in 35 community-dwelling men over the age of 75 years and only a weak inverse correlation between total T and spine BMD that lost significance after adjustment for BMI. In the present study, serum DHEAS levels were inversely associated with BMD of the trochanter. Aromatization of androgens to oestrogens may explain, at least in part, the effects of androgens on skeletal maintenance (Vanderschueren et al., 2000). Falahati-Nini et al. (2000) lowered testosterone levels in older men and then replaced either oestradiol or testosterone and showed that oestradiol inhibited osteoclast and promoted osteoblast activity, whereas testosterone promoted osteoblast activity but had no effect on bone resorption by osteoclasts. A post hoc analysis of the association between E2 and BMD of the trochanter (R² = −0.27, p = 0.035) in this cohort revealed an almost-identical inverse association to that between DHEAS and trochanter BMD. Conversely, in a large cross-sectional study of 437 elderly men, low oestradiol and increased SHBG were both independent predictors of low BMD at the lumbar spine and hip, and low FT was an independent predictor of low BMD at the spine but not at the hip (Center et al. 1999). It is unclear why the weak associations between BMD and DHEAS and E2 are negative in this cohort. As osteoporotic fracture is increasingly becoming a problem in men, as it is in women, BMD should be measured in men who have been identified as hypogonadal.

In the present study, there was a lack of association between the androgen levels studied and depressive symptoms as measured by CESD scores. Hypogonadal men have been shown to be more depressed, angered, fatigued and confused than infertile, treated eugonadal or normal men (Barrett-Connor et al., 1999) and positive relationships between androgen levels and mood and well-being have been reported (Yesavage et al., 1985). BT has been reported to be 17% lower in categorically defined depressed men than in normal healthy men (Barrett-Connor et al., 1999).

Debate continues over whether measurement of BT provides additional information over and above that of total T. BT assays are increasingly used in North America. Wide variance in total T between laboratories using different platform assays have been reported, with some platforms performing more reliably than others (Wang et al., 2004). This ultimately transfers to variance in BT concentrations as total T is required to calculate BT concentration from the per cent BT obtained from the ammonium sulphate precipitation. A number of calculated FT and BT methods are currently in use that also employ the measurement of total T and SHBG (Vermeulen et al., 1999; Ly & Handelsman, 2005). It is thus, critical to accurately and reliably measure total T.

This study was limited by the fact that blood was sampled at various times of the day and was not controlled for post-absorptive state. This could partially explain the lack of age-related decline in serum T and BT levels in this group as not all men were sampled when their levels were most likely to reflect their circadian peak. Moreover, triglyceride and glucose levels are subject to prior dietary intake and may lead to overestimation of the true fasting levels. Moreover, this was a highly exploratory study using cross-sectional analyses in a relatively small group of men and many statistical tests were performed in order to arrive at the final models. Thus, some false-positive correlations are likely. It is also possible that the relatively few positive findings in the present study are due to a lack of statistical power to detect weak but biologically significant associations. Moreover, there is potential in this study for some degree of selection bias, in that 124 of 371 men in the original sample (33%) chose to participate in special studies from which these data were derived. The novelty, however, of this study is that ageing African-American men are a highly understudied group, and to date, few data exist on functional associations with circulating androgens in this population.

Taken together, in the context of the present cross-sectional study of middle-aged African-American men, total T and BT levels remain relatively constant between the ages of 50 and 65 years, whereas DHEAS shows a significant decline. Total T levels appeared to be associated (albeit weakly) with serum biochemical markers of metabolic status, whereas BT was associated with muscle and cognitive functioning. DHEAS levels were associated inversely with adiponectin levels and positively with MMSE scores. Sociodemographic and lifestyle factors may modify associations of biochemical and functional variables with serum androgen levels.